Spatial prediction of landslide susceptibility/intensity through advanced statistical approaches implementation: applications to the Cinque Terre (Eastern Liguria, Italy)

Landslides are frequently responsible for considerable huge economic losses and casualties in mountainous regions especially nowadays as development expands into unstable hillslope areas under the pressures of increasing population size and urbanization (Di Martire et al. 2012). People are not the only vulnerable targets of landslides. Indeed, mass movements can easily lay waste to everything in their path, threatening human properties, infrastructures and natural environments. Italy is severely affected by landslide phenomena and it is one of the most European countries affected by this kind of phenomena. In this framework, Italy is particularly concerned with forecasting landslide effects (Calcaterra et al. 2003b), in compliance with the National Law n. 267/98, enforced after the devastating landslide event of Sarno (Campania, Southern Italy). According to the latest Superior Institute for the Environmental Protection and Research (ISPRA, 2018) report on "hydrogeological instability" of 2018, it emerges that the population exposed to landslides risk is more than 5 million and in particular almost half-million falls into very high hazard zones. The slope stability can be compromised by both natural and human-caused changes in the environment. The main reasons can be summarised into heavy rainfalls, earthquakes, rapid snow-melts, slope cut due to erosions, and variation in groundwater levels for the natural cases whilst slopes steepening through construction, quarrying, building of houses, and farming along the foot of mountainous zone correspond to the human component. This Ph.D. thesis was carried out in the Liguria region, inside the Cinque Terre National Park. This area was chosen due to its abundance of different types of landslides and its geological, geomorphological and urban characteristics. The Cinque Terre area can be considered as one of the most representative examples of human-modified landscape. Starting from the early centuries of the Middle Ages, local farmers have almost completely modified the original slope topography through the construction of dry-stone walls, creating an outstanding terraced coastal landscape (Terranova 1984, 1989; Terranova et al. 2006; Brandolini 2017). This territory is extremely dynamic since it is characterized by a complex geological and geomorphological setting, where many surficial geomorphic processes coexist, along with peculiar weather conditions (Cevasco et al. 2015). For this reason, part of this research focused on analyzing the disaster that hit the Cinque Terre on October, 25th, 2011. Multiple landslides took place in this occasion, triggering almost simultaneously hundreds of shallow landslides in the time-lapse of 5-6 hours, causing 13 victims, and severe structural and economic damage (Cevasco et al. 2012; D’Amato Avanzi et al. 2013). Moreover, this artificial landscape experienced important land-use changes over the last century (Cevasco et al. 2014; Brandolini 2017), mostly related to the abandonment of agricultural activity. It is known that terraced landscapes, when no longer properly maintained, become more prone to erosion processes and mass movements (Lesschen et al. 2008; Brandolini et al. 2018a; Moreno-de-las-Heras et al. 2019; Seeger et al. 2019). Within the context of slope instability, the international community has been focusing for the last decade on recognising the landslide susceptibility/hazard of a given area of interest. Landslide susceptibility predicts "where" landslides are likely to occur, whereas, landslide hazard evaluates future spatial and temporal mass movement occurrence (Guzzetti et al., 1999). Although both definitions are incorrectly used as interchangeable. Such a recognition phase becomes crucial for land use planning activities aimed at the protection of people and infrastructures. In fact, only with proper risk assessment governments, regional institutions, and municipalities can prepare the appropriate countermeasures at different scales. Thus, landslide susceptibility is the keystone of a long chain of procedures that are actively implemented to manage landslide risk at all levels, especially in vulnerable areas such as Liguria. The methods implemented in this dissertation have the overall objective of evaluating advanced algorithms for modeling landslide susceptibility. The thesis has been structured in six chapters. The first chapter introduces and motivates the work conducted in the three years of the project by including information about the research objectives. The second chapter gives the basic concepts related to landslides, definition, classification and causes, landslide inventory, along with the derived products: susceptibility, hazard and risk zoning, with particular attention to the evaluation of landslide susceptibility. The objective of the third chapter is to define the different methodologies, algorithms and procedures applied during the research activity. The fourth chapter deals with the geographical, geological and geomorphological features of the study area. The fifth chapter provides information about the results of the applied methodologies to the study area: Machine Learning algorithms, runout method and Bayesian approach. Furthermore, critical discussions on the outcomes obtained are also described. The sixth chapter deals with the discussions and the conclusions of this research, critically analysing the role of such work in the general panorama of the scientific community and illustrating the possible future perspectives.